解决Oracle RMAN删除归档日志不释放问题的方法
前言
当Oracle 归档日志满了后，将无法正常登入oracle，需要删除一部分归档日志才能正常登入ORACLE。
最近同事在工作中遇到一个问题，他那边的一套Oracle 11g数据库使用RMAN没办法删除旧的归档，导致磁盘使用率很高。
oracle@ps251n2s:[/opt/oracle/archive/db] ls -ltr |more total 3059881082 -rw-r----- 1 oracle oinstall 725320704 May 20 2016 1_1_91233774.dbf -rw-rw---- 1 oracle oinstall 1756937216 Jun 7 2016 1_2_91233774.dbf -rw-rw---- 1 oracle oinstall 1756731392 Jun 21 2016 1_3_91233774.dbf -rw-rw---- 1 oracle oinstall 1758001152 Jul 3 2016 1_4_91233774.dbf -rw-rw---- 1 oracle oinstall 1756729344 Jul 15 2016 1_5_91233774.dbf -rw-rw---- 1 oracle oinstall 1765750784 Jul 26 22:00 1_6_91233774.dbf -rw-rw---- 1 oracle oinstall 1760346112 Aug 6 14:04 1_7_91233774.dbf ......
使用RMAN list archivelog all;查看日志也没有显示旧归档日志的记录
RMAN> list archivelog all; using target database control file instead of recovery catalog List of Archived Log Copies for database with db_unique_name DBS ===================================================================== Key Thrd Seq S Low Time ------- ---- ------- - --------- 1206 1 1206 A 14-JAN-17 Name: /opt/oracle/archive/db/1_1206_91233774.dbf 1207 1 1207 A 14-JAN-17 Name: /opt/oracle/archive/db/1_1207_91233774.dbf 1208 1 1208 A 14-JAN-17 Name: /opt/oracle/archive/db/1_1208_91233774.dbf ......
使用crosscheck检查也没有查到旧归档日志记录
RMAN> crosscheck archivelog all; using target database control file instead of recovery catalog allocated channel: ORA_DISK_1 channel ORA_DISK_1: SID=16 device type=DISK validation succeeded for archived log archived log file name=/opt/oracle/archive/db/1_1206_91233774.dbf RECID=1206 STAMP=933244655 validation succeeded for archived log archived log file name=/opt/oracle/archive/db/1_1207_91233774.dbf RECID=1207 STAMP=933249510 validation succeeded for archived log archived log file name=/opt/oracle/archive/db/1_1208_91233774.dbf RECID=1208 STAMP=933254500 validation succeeded for archived log archived log file name=/opt/oracle/archive/db/1_1209_91233774.dbf RECID=1209 STAMP=933259445 validation succeeded for archived log
查看v$archived_log视图也没有旧归档日志记录
解决过程
于是在网上找到一篇文章，大体意思是MAN删除归档日志的依据是controlfile中记录的归档日志，而不是你磁盘上实际存在的归档日志，如果你的归档日志已经不在controlfile中就无法被RMAN识别到，也就无法被删除。而致于归档信息何时在控制文件中被覆盖，这由一个参数control_file_record_keep_time来控制。
查看数据库中的参数是默认的7天
SYS@db> show parameter CONTROL_FILE_RECORD_KEEP_TIME NAME TYPE VALUE ------------------------------------ -------------------------------- ------------------------------ control_file_record_keep_time integer 7
难道是在控制文件中归档日志的记录条数只能保留7天的么？于是到MOS去找相关文档找到一篇Why Are Controlfile Entries For Archived Logs Not Overwritten After CONTROL_FILE_RECORD_KEEP_TIME? (文档 ID 1056085.1)，文档介绍CONTROL_FILE_RECORD_KEEP_TIME参数并不是强制在控制文件中保留多少天的归档日志条目，而是由V$CONTROLFILE_RECORD_SECTION视图中的ARCHIVED LOG条目去控制的，当RECORDS_TOTAL=RECORDS_USED时旧的归档日志就会被覆盖掉。
于是查数据库中的V$CONTROLFILE_RECORD_SECTION视图，看到ARCHIVED LOG行的records_total=224意味着控制文件中只能保留224条归档日志文件信息。
SYS@dbs> select * from v$controlfile_record_section where type='ARCHIVED LOG'; TYPE RECORD_SIZE RECORDS_TOTAL RECORDS_USED FIRST_INDEX LAST_INDEX LAST_RECID ---------------------------- ----------- ------------- ------------ ----------- ---------- ---------- ARCHIVED LOG 584 224 224 88 87 1235
查询v$archived_log视图，总数为224条
SYS@dbs> select count(*) from v$archived_log; COUNT(*) ---------- 224
这时应该可以推断出由于控制文件中只能保留224条归档日志数据，所以以前的归档信息已经被控制文件刷出，RMAN无法识别到这些归档文件，所以就无法删除这些归档日志。
那么现在如果使用RMAN删除这些旧的归档日志呢？我想到使用catalog start with 命令重新注册这些文件到控制文件中
RMAN> catalog start with '/opt/oracle/archive/db'; ...... File Name: /opt/oracle/archive/db/1_1005_91233774.dbf File Name: /opt/oracle/archive/db/1_1006_91233774.dbf File Name: /opt/oracle/archive/db/1_1007_91233774.dbf Do you really want to catalog the above files (enter YES or NO)? yes cataloging files... cataloging done ...... RMAN> list archivelog all; ..... 1236 1 301 A 22-OCT-16 Name: /opt/oracle/archive/db/1_301_91233774.dbf 1237 1 302 A 22-OCT-16 Name: /opt/oracle/archive/db/1_302_91233774.dbf ......
旧的归档日志文件已经注册到控制文件中了，再使用RMAN进行删除就可以了
delete archivelog until logseq 1007;
再次查询v$controlfile_record_section视图，records_total也被“撑”大了
SYS@db> select * from v$controlfile_record_section where type='ARCHIVED LOG'; TYPE RECORD_SIZE RECORDS_TOTAL RECORDS_USED FIRST_INDEX LAST_INDEX LAST_RECID ---------------------------- ----------- ------------- ------------ ----------- ---------- ---------- ARCHIVED LOG 584 896 742 807 652 1800
至此问题结束。
总结